DP Why is Exporting Hard in Some Sectors? Anders AKERMAN

DP
RIETI Discussion Paper Series 13-E-015
Why is Exporting Hard in Some Sectors?
Anders AKERMAN
Stockholm University
Rikard FORSLID
Stockholm University
OKUBO Toshihiro
Keio Univesity
The Research Institute of Economy, Trade and Industry
http://www.rieti.go.jp/en/
RIETI Discussion Paper Series 13-E-015
March 2013
Why is Exporting Hard in Some Sectors? *
Anders AKERMAN †
Stockholm University
Rikard FORSLID ‡
Stockholm University
OKUBO Toshihiro
§
Keio University
Abstract
This paper models the market entry cost of exporters as dependent on the size of the export market
as well as on sector specific factors. We introduce these features in a Melitz trade model with
heterogeneous firms. The predictions of our model are tested using Swedish and Japanese firm level
data. We find that sector level advertising or sales promotion intensity is an important component of
the market entry cost. A larger market size, if anything, lowers entry costs.
Keywords: Heterogeneous firms, Market size, Market entry costs
JEL classification: D21, F12, F15
RIETI Discussion Papers Series aims at widely disseminating research results in the form of professional
papers, thereby stimulating lively discussion. The views expressed in the papers are solely those of the
author(s), and do not represent those of the Research Institute of Economy, Trade and Industry.
*
Financial support from the Swedish Research Council and Handelsbanken research foundation is gratefully
acknowledged by Akerman and Forslid.
† Stockholm University, email: anders.akerman@ne.su.se.
‡ Stockholm University, CEPR; email: rf@ne.su.se.
§ Keio University; email: okubo@econ.keio.ac.jp
This study is conducted as a part of the project “Study of the Creation of the Japanese Economy and Trade and Direct Investment”
undertaken at Research Institute of Economy, Trade and Industry (RIETI).
1
Introduction
It is empirically well established that exporting …rms tend to be more productive, larger, and
endure longer than domestic …rms; see Tybout (2003) for a survey.1 One of the proposed reasons
for this is the costs of entering a foreign market. The purpose of this paper is to investigate the
nature of market entry costs. We employ a heterogeneous …rms and trade model a la Melitz
(2003), where the market entry cost is modelled as having a …xed and a variable part as well as
a sector speci…c component.
While the standard models of heterogeneous …rms employ a …xed and exogenous market
entry cost, Arkolakis (2010) proposes a model with an endogenous market entry cost. He uses
a model of advertising where the probability that an advertising message is received by one
consumer increases with the size of the population. Thus, a …rm gets a relatively larger payo¤
for a small investment in marketing when exporting to larger countries. The marginal exporter
that is just su¢ ciently productive to export will therefore prefer to export to a large rather
than to a small market. A small …rm will therefore …nd it easier to enter a large market, and
this is consistent with the empirical observation by e.g. Eaton, Kortum, and Kramarz (2005)
that show that many small …rms typically export a small amount to large markets.
As in Arkolakis (2010), we allow the market size to a¤ect entry costs, but while a large
market may be easier to enter because it is easier to …nd consumers that match a particular
product, it may also be the case that it is more di¢ cult to enter a large market e.g. due to
higher advertising requirements. Our speci…cation allows for both these possibilities. We also
model the market entry cost as a product group or sector dependent. It may e.g. require more
marketing to establish a branded consumer good, such as soft drinks or clothes, in a foreign
market than to establish a more standardised product.
We test the predictions of our model using data on Swedish …rm level export to Japan
combined with Japanese …rm level data. We use the Japanese data to calculate market speci…c
advertising outlays, sales promotion and …rm sales. We also have sector level data for tari¤s.
Our empirical part support the notion that higher average adverting outlays at the sector level
makes it more di¢ cult to enter a foreign market. A larger market size, if anything, makes it
easier to enter a market. However, this e¤ect is not signi…cant in the regressions.
The paper is organized as follows: Section 2 contains the model. Section 3 contains empirical
tests. Finally, section 4 concludes the paper.
2
The Model
This paper introduces a formulation of the market entry cost that is sector and market speci…c in
a multisector version of Melitz (2003) monopolistic competition trade model with heterogeneous
…rms.
1
Other studies include Aw, Chung, and Roberts (2000), Bernard and Jensen (1995, 1999a, 1999b) Clerides,
Lach, and Tybout (1998) as well as Eaton, Kortum, and Kramarz (2004).
1
2.1
Basics
There are two countries (Home and Foreign) and multiple sectors. The A-sector (traditional
sector) is a Walrasian, homogeneous-goods sector with costless trade. There is also a continuum
of M-sectors (manufactures) marked by increasing returns, Dixit-Stiglitz monopolistic competition and iceberg trade costs. M-sector …rms face constant marginal production costs and three
types of …xed costs. Each country has a single primary factor of production labour, L, used
in each sector. For simplicity, there is no comparative advantage across the M-sectors. All
manufacturing sectors have an identical structure but …nal demand for each sector di¤ers.
Firms in the manufacturing sectors are heterogeneous a la Melitz(2003). In essence, there is
heterogeneity with respect to …rms’marginal costs. Each Dixit-Stiglitz …rm/variety is associated
with a particular labour input coe¢ cient – denoted as ai for …rm i. After sinking FE units of
labour in the product innovation process, the …rm is randomly assigned an ‘ai ’from a probability
distribution G(a). On the other hand there are three types of …xed costs. The …rst …xed cost,
FE , is the standard Dixit-Stiglitz cost of developing a new variety. The second and third …xed
costs are ‘beachhead’costs re‡ecting the one-time expense of introducing a new variety into a
market. These costs are here assumed to depend on the size of the market.
Consumers in each nation have two-tier utility functions with the upper tier (Cobb-Douglas)
determining the consumer’s division of expenditure among the sectors and the second tier (CES),
dictating the consumer’s preferences over the various di¤erentiated varieties within each Msector.
All individuals have the utility function
Uj =
0 ln CA
+
is the set of manufacturing sectors,
Z
Z
j
ln CM j dj
(1)
j2
where
j dj
=1
0,
and CA is consumption of the
j2
homogenous good. Each manufacturing sector j enters the utility function through the index
CM j ; de…ned by
CM j
2
ZNj
6
(
= 4 cij
1)=
0
3
7
di5
=(
1)
;
(2)
Nj being the mass of varieties consumed in sector j, cij the amount of variety i consumed in
sector j; and
> 1 the elasticity of substitution. Each consumer spends a share
j
of his income
on sector j, and demand for a variety i in sector j is therefore
xij =
pij
Pj1
j Y;
where pij is the consumer price of variety i in sector j, Y is income; and Pj
(3)
N
Rj
0
the price index of manufacturing goods in sector j.
2
p1ij
!
di
1
1
The unit factor requirement of the homogeneous good (Agriculture) is one unit of labour.
This good is freely traded and since it is chosen as the numeraire
pA = w = 1;
(4)
w being the nominal wage of workers in all countries.
Shipping the manufactured good involves a frictional trade cost of the “iceberg” form: for
one unit of a good from one country to the other,
> 1 units must be shipped. It is assumed that
trade costs are equal in both directions and that
= 1: Pro…t maximization by a manufacturing
…rm i leads to consumer price pi =
1
ai .
Manufacturing …rms draw their marginal cost, a; from the probability distribution G(a) after
having sunk FE units of labour to develop a new variety. Having learned their productivity,
…rms decide on entry in the domestic and foreign market, respectively. Firms will enter a market
as long as the operating pro…t in this market is su¢ ciently large to cover the beachhead (market
entry) cost associated with the market. Because of the constant mark-up pricing, it is easily
shown that operating pro…ts equal sales divided by . Using this and (3), the critical ’cut-o¤’
levels of the marginal costs in sector j are given by:
1
Bj = Fd (Dj );
aDj
1
aXj
(5)
Bj = Fx (Dj );
(6)
where Dj (Dj ) is Home (Foreign) demand in sector j, i.e. Dj
represents trade freeness. Bj
Dj
j
jL
1
and
represents market potential where
j
(
1 1
)
2 [0; 1]
P1
.
The market entry cost (beachhead cost) is assumed to depend on the demand size, and we will
parametrize this dependence below. However, note that it is natural that the beachhead costs
F s are sector speci…c. For example, branded consumer goods may require more marketing than
raw materials.
Finally, free entry ensures that the ex-ante expected pro…t of developing a new variety in
sector j equals the investment cost:
aDj
Z
0
2.2
a1
Bj
Fd (Dj ) dG(a) +
aXj
Z
a1
Bj
Fx (Dj ) dG(a) = FE :
(7)
0
Solving for the Long-run Equilibrium
First, our model is two-country with multiple sectors but the model is solvable in terms of a
representative sector since the sector level expenditure shares,
j;
are …xed . We focus on one of
Manufacturing sectors. Without loss of generality we drop subscript j, since sectors are mirror
3
images. Second, we follow Helpman, Melitz, and Yeaple (2004) in assuming the probability
density function to be a Pareto distribution, which a cumulative density function given by2 :
a
a0
G(a) =
where
:
(8)
is shape parameter and the scale parameter, a0 , is normallised as unity. Substituting
the cut-o¤ conditions (5) and (6) into the free-entry condition (7) gives B,
B=
where
1
> 1, and
FE Fd
1
1
(D) (
1) (1
1
(D) (D )
Fx (D)
Fd (D)
(D)
1
(D ))
!1
;
(9)
2 [0; 1] is a measure of openness (See Baldwin
and Forslid (2010)). Using (9) and the cut-o¤ conditions gives the cut-o¤ marginal costs:
aD =
aX =
(
(
(1
1) FE
Fd (D)
(D ))
(D) (D )
1
1) (D )FE
Fx (D )
(1
;
(10)
(D))
(D) (D )
1
:
(11)
From these, it is seen that, contrary to the standard model by Melitz (2003), the market size
will a¤ect the cut-o¤ marginal costs. We will assume that FX (D )
1
(D )
1
(D)
1
> FD (D) for all
sectors j:3 This assumption implies that aX < aD .
may be written as
=
1
1
naD
(1
+n
aD
)
aX
aD
+1
!
;
(12)
and the mass of …rms in each country can be calculated using (9), (10), and (11) together with
the fact that B =
D
:
n=
(
1) D (1
Fd (D)
(1
(D)) D (D) (1
(D ))
:
(D) (D )) (1
(D))
(13)
Welfare may be measured by indirect utility, which is proportional to the real wage
It su¢ ces to examine Pj . Using (10), (11), (12), and (13), we have
P =(
1
) D
Fd
1
(D)FE (
1)
1
1
(D )
(D) (D )
R wj
ln P
:
1
(
1)
:
(14)
This expression shows that, as in the Melitz (2003) model, welfare always increases (P decreases)
with trade liberalization; that is, with a higher
or a lower
FX (D )
FD (D) .
2
This assumption is consistent with the empirical …ndings by e.g. Axtell (2001).
3
The corresponding condition in Melitz (2003) is that
4
FX
> FD :
2.3
Parametrisation of the market entry cost
In the following, we parametrise the sector level market entry costs as:
Fdj (D)
j
fd + Dj ;
Fxj (D )
fx + Dj
j
;
j
> 0;
7 0:
(15)
The variable component of the beachhead cost is dependent on market size, while the constant
terms picks up costs that are independent of market size. It is quite natural that the beachhead
cost would have one …xed and one variable component. The constant f represents the …xed cost
of standardizing a product for a particular market or the cost of producing an advertisement
tailored to a particular market with its culture and language. We assume that these costs are
larger when a …rm enters a foreign market: fX > fD : The variable cost term Dj represents the
fact that the cost of entry depends on the demand (market size), Dj ; in each sector j.
>0
implies that the cost of entry increases in the size of the market. For instance, because the
number of free product samples or advertising posters increases in the size of the market.
<0
implies that the cost of entry decreases in the size of the market. This may be e.g. because
it is more likely that at least some consumers buy your product in a large market (Arkolakis
(2010)). The sign of
may depend on the type of product, but as in previous studies, data only
allows us to identify the sector level average
Finally, the sector speci…c parameter
j
in the empirical part below.
represents the possibility that market entry costs
di¤er among di¤erent product groups or sectors. Branded consumer goods may e.g. require
more more marketing than standardised inputs or raw materials.
2.4
Export cuto¤s and the foreign market
The cut-o¤ marginal cost of an exporter in sector j is from (11) given by
aXj =
(
0
1) FE @ (1
1
Fx (Dj )
(D
j)
1
(Dj )) A
:
(Dj )
(16)
Substituting in the market entry cost for the foreign market from (15) gives
aXj =
(
j
1) FE
fx + Dj
0
B
B
B
@
(1
fx +Dj
fd +Dj
1
(Dj ))
1
(Dj )
C
C
C:
A
(17)
A …rst result, which can be seen directly from the expression, implies that the export cut-o¤
is tougher in markets with a high
j.
Proposition 1 The cut-o¤ productivity of exporters increases in
5
j:
Proof: The result follows directly from (17).
The e¤ect of the foreign market size, , D ; is given by the following proposition
Proposition 2 The cut-o¤ productivity of exporters increases in D if
D if
> 0 and decreases in
< 0 given that fX > fD :
Proof: See Appendix.
We now turn to data to test these predictions.
3
Empirics
Our theory should in principle be tested using cross country …rm level data. However, such
data is hard to …nd. There would in any case be a problem of unobserved heterogeneity among
countries that makes it di¢ cult to establish the source of varying entry costs. We therefore
here use Swedish …rm level export to one country, where each sector is considered a separate
market. We use Swedish export to Japan combining Swedish …rm level data with Japanese data.
Sweden is one of the most successfully export countries in the world, and most Swedish …rms
are exporters. On the other hand, Japan is one of the most closed economies, with a market
that is di¢ cult to penetrate for foreign …rms. Furthermore, there is a substantial physical and
cultural distance between Sweden and Japan, and we therefore expect …xed market entry costs
to be relatively important for Swedish exports to Japan.
3.1
Data
We use manufacturing census data for Sweden. The census data contain information on a large
number of variables such as export (tSEK), employment (number of employees), capital stock
(tSEK), use of raw materials (tSEK), value added and production value (tSEK) at the …rm
level from 2000-2007. We can therefore calculate …rms level productivity as the residual from
sector-speci…c production functions estimated by the Levinson Petrin method.
We have information about the Japanese market from two data sources. One is sectoral
aggregation from …rm level data, entitled the Results of the Basic Survey of Japanese Business
Structure and Activities (Kigyou Katsudo Kihon Chosa in Japanese) of year 2005 from Ministry
of Economy, Trade and Industry of Japan (METI). This dataset provides information on the
activities and organisation for all Japanese …rms with more than 50 employees and a capital
of more than 30 million yen. The sample includes not only Japanese …rms (i.e. domestic
…rms) but also foreign owned …rms, which is de…ned as …rms with a foreign ownership of more
than 33 percent. We aggregate this …rm level data to the 3-digit sector level. The other data
source is the Japanese Industrial Productivity Database (JIP data) prepared by the Research
Instutute of Economy, Trade and Industry (RIETI). This dataset contains sectoral information
about the Japanese market (108 sectors in agriculture, mining, manufacturing and service). We
calculate the sales promotion ratio and the advertising cost ratio at the sectoral level, using
6
sales promotion costs, advertisement expenditure and total sales from the Results of the Basic
Survey of Japanese Business Structure and Activities dataset. It is possible that the advertising
and marketing requirements di¤er among foreign and domestic …rms in Japan, and we therefore
both use the average sales promotion of foreign …rms in Japan by sector and a measure of
advertising per sales for all …rms. From the JIP data we use data on demand and tari¤ rates
at the sectorial level. Finally, we match these sectoral variables to the Swedish data.
3.2
Results
Consider …rst the relationship between the marketing intensity in a market and exporter productivity. We use the advertising ratio (advertising per sales) as well as the sales promotion
ration (sales promotion per sales) as measures of marketing intensity. The main di¤erence being that the sales promotion data is collected only among foreign …rms in Japan, whereas the
advertising ratio is for all …rms in the market. We have estimated …rm productivity using the
Levinson Petrin method. To have an impression of the importance of the marketing intensity
of a market for the productivity of exporters we divide the markets (sectors) into two groups:
those above and those below the mean advertising ratio (advertising per sales) and those above
and below the mean sales promotion ratio. We then make a kernel density plot for exporters
belonging to the two groups of exporter. Figure 1 shows how the entire productivity distribution of exporters to markets with intense advertising or sales promotion is shifted to the right
compared to exporters to the less marketing intensive markets. This pattern is present for both
measures of marketing costs.
7
0
Kernel density
.2
.4
.6
.8
Adv ertisement ratio
2
3
4
5
6
7
5
6
7
0
Kernel density
.2
.4
.6
.8
Sales promotion ratio
2
3
4
Below median
Abov e m edian
Figure 1: Productivity distributions of Swedish exporters to Japan.
Turning to sector level averages, Figure 2 and Figure 3 plot our two measures of marketing
intensity against sector average exporter productivity. Both …gures show a marked positive
relationship between exporter productivity and the market intensity in a market. A very similar
pattern emerges when plotting productivity calculated at the 3-digit sector level as shown in
Appendix. The pattern of Figure 2 and Figure 3 is consistent with Proposition 1: the exporting
cut-o¤ productivity being higher in markets with a higher marketing intensity (
8
j ).
5
1
2
Productivity (log)
3
4
5
4
Productivity (log)
3
2
1
-10
-8
-6
Advertising exp./sales (log)
-4
-2
-2
-1.5
Sales promotion exp./sales (log)
-1
Figure 2: Advertising intensity and exporter
Figure 3: Sales promotion intensity and exporter
productivity.
productivity.
We turn to next to the e¤ect of the size of the export market. Figure 4 plots the size
1
Productivity (log)
2
3
4
5
(demand) of Japanese sectors against the average productivity of the exporters.
8
10
12
Market size (demand) (log)
14
16
Figure 4: Market size and exporter productivity.
Figure 4 shows a negative relationship between average productivity and market size (a
negative ): That is, this pattern is consistent with entry costs being lower in large markets,
as proposed by Arkolakis (2010). However, the relationship between exporter productivity and
market size will not turn out signi…cant in the regressions below.
Finally, Table 1 shows a set of regressions with exporter productivity as the dependent
9
variable. We use sector level demand and average sector level tari¤s as controls alongside with
our two measures of marketing intensity. We show regressions with and without clustering
at the sector level, as well as one regression where the data has been collapsed by sector. In
particular sales promotion per sales by foreign …rms in Japan has a very strong and signi…cant
e¤ect on average exporter productivity. Also advertising per sales averaged over all …rms in a
sector has a signi…cant e¤ect in two of the three regressions. The point estimate at the sectoral
regression in (3) is larger than in the …rm level regression but we lack the precision to estimate
it with statistical signi…cance.
Table 1: Regressions results
Dependent variable:
Advertisement ratio (log)
Productivity (log)
(1)
(2)
(3)
0.110***
0.110*
0.177
(0.0183)
(0.0593)
(0.140)
Sales promotion ratio (log)
Demand (log)
(4)
(5)
(6)
0.476***
0.476*
1.333**
(0.0812)
(0.255)
(0.555)
-0.0275
-0.0275
-0.0204
-0.0113
-0.0113
0.0493
(0.0170)
(0.0734)
(0.157)
(0.0180)
(0.0692)
(0.144)
0.128
0.128
-0.0975
0.178**
0.178
-0.195
(0.0833)
(0.299)
(0.586)
(0.0817)
(0.279)
(0.508)
Observations
785
785
18
785
785
18
R-squared
0.06
0.06
0.115
0.06
0.06
0.30
Clustered standard errors
NO
YES
NO
NO
YES
NO
Tari¤ (log)
Note: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.
4
Conclusion
This paper has introduced a market and sector dependent market entry or beachhead cost in the
heterogeneous …rms and trade model of Melitz (2003). We model this cost as having one variable
component that depends on the market size, and one …xed component. The …xed component
could e.g. be interpreted as the cost of standardizing a product for a particular market, while
the variable cost term e.g. represents the fact that the advertising cost of introducing a new
product may depend on the size of the market (the number of consumers). We do not put any
restriction on the e¤ect of the market size. A larger market may be easier to enter because it
is easier to …nd consumers for a particular product, or it may be harder to enter due to higher
costs of e.g. sampling. We also introduce a sector speci…c component in the formulation of the
market entry cost. This component picks up sector level di¤erences e.g. in advertising intensity.
10
In essence we make a relatively small change to the framework of Helpman, Melitz, and Yeaple
(2004), which preserves the analytical solvability of the model.
We combine Swedish …rm level data with Japanese data to test the predictions of the model.
The empirical part suggests that exporting is harder when there is a high advertising intensity
in the export market. If anything, we …nd a negative relationship between exporter productivity
and size of the export market, but this e¤ect is not signi…cant in the regressions.
11
References
Arkolakis, C. (2010): “Market Penetration Costs and the New Consumers Margin in International Trade,” Journal of Political Economy, 118(6), 1151–1199.
Aw, B., S. Chung, and M. Roberts (2000): “Productivity and Turnover in the Export
Market: Micro Evidence from Taiwan and South Korea,” World Bank Economic Review,
14(6), 65–90.
Axtell, R. L. (2001): “Zipf Distribution of U.S. Firm Sizes,” Science, 293(6), 1818–1820.
Baldwin, R. E., and R. Forslid (2010): “Trade liberalisation with heterogeneous …rms,”
Review of Development Economics, 14(2), 161–176.
Bernard, A. B., and J. B. Jensen (1995): “Exporters, Jobs and Wages in U.S. Manufacturing, 1976-1987,” Brookings Papers on Economic Activity, Microeconomics, 14(6), 67–119.
(1999a): “Exceptional Exporter Performance: Cause, E¤ect, or Both?,” Journal of
International Economics, 47(6), 1–26.
(1999b): “Exporting and Productivity: Importance of Reallocation,” NBER Working
Paper, 7135(6), 67–119.
Clerides, S. K., S. Lach, and J. R. Tybout (1998): “Is Learning by Exporting Important? Microdynamic Evidence from Colombia, Mexico and Morocco,” Quarterly Journal of
Economics, 113(3), 903–947.
Eaton, J., S. Kortum, and F. Kramarz (2004): “Dissecting Trade: Firms, Industries, and
Export Destinations,” NBER Working Papers, 10344(3), 150–154.
(2005): “An Anatomy of International Trade: Evidence from French Firms.,” 2005
Meeting Papers, Society for Economic Dynamics, 197.
Helpman, E., M. J. Melitz, and S. R. Yeaple (2004): “Export versus FDI with Heterogeneous Firms,” American Economic Review, 94(4), 300–316.
Melitz, M. J. (2003): “The Impact of Trade on Intra-Industry Reallocations and Aggregate
Industry Productivity,” Econometrica, 71(4), 1695–1725.
Tybout, J. (2003): “Plant- and Firm-level Evidence on "New" Trade Theories,”in E. K. Choi
and J. Harrigan (eds) Handbook of International Trade Blackwell, Oxford.
12
5
Appendix
5.1
Proof of proposition 2
Taking logs of (11) gives
1
ln aXj = fln((
1) FE )(1
(Dj ))
ln(Fx (Dj ))
ln(
1
(Dj )
(Dj ))g:
Di¤erentiation of ln aX in terms of D gives
d ln aX
=
dD
1
dFx (D )
+
Fx (Dj ) dD
(
1
1
(Dj )
1
(Dj )
(Dj ))
!2
d (D )
:
dD
To sign this expression note …rst that
dFx (D )
dD
dFx (D )
dD
> 0 for
>0
< 0 for
< 0;
from the de…nition of Fx :
Next, to determine the sign of
ln (D ) = (1
d (D )
dD
we take the logarithm of
) ln(fx + Dj )
(1
(D ):
) ln(fd + Dj ) +
ln :
Di¤erentiation w.r.t. D gives
d ln (D )
= (1
dD
) D
1
1
fx + Dj
Given the assumption that fx > fd we have:
d ln (D )
dD
d ln (D )
dD
< 0 for
>0
> 0 for
<0
Combining these results gives
d ln aX
dD
d ln aX
dD
< 0 for
>0
> 0 for
< 0:
13
1
fd + Dj
!
4
Productivity (log)
3
2
1
1
2
Productivity (log)
3
4
5
Exporter productivity and the intensity of marketing
5
5.2
-10
-8
-6
Advertising exp./sales (log)
-4
-2
-2.5
-2
-1.5
Sales promotion exp./sales (log)
-1
Figure 2b: Advertising intensity and exporter
Figure 3b: Sales promotion intensity and
productivity at the 3-digit level.
exporter productivity at the 3-digit level.
14
-.5